Institute for Simulation
UCF College of Medicine
IST medical simulation page
Photos: Contextual Anatomic Mimesis (ISMAR 2007) Christoph Bichlmeier, TUM Camp Lab, Munich Germany
Lectures in the Series
Please join IST, the National Center for Simulation and UCF's new College of Medicine at our lecture series intended to stimulate academic discussion on medical training. During the series we hope to explain current and introduce emerging technology designed to enhance that training.
Cognitive Simulation: The Maryland Virtual Patient (MVP)
Sergei Nirenburg, Marjorie McShane, Stephen Beale
University of Maryland, Baltimore County
Bruce Jarrell, George Fantry
University of Maryland School of Medicine
Monday, March 16, 2009
A 45-minute presentation will be followed by 15 minutes of discussion
The Maryland Virtual Patient (MVP) is an electronic environment designed for trainees to rehearse and perfect the mental aspects of medical decision-making. This simulated environment aims to recreate the real life medical situation that is among our most effective learning situations; direct patient care in a teaching hospital or clinic. In this
situation, a trainee communicates with a patient, interprets test results and images, establishes a diagnosis, develops a plan with the patient, performs interventions and observes responses to treatments, all while under the supervision of a clinician-teacher. Throughout the process, sophisticated language-based exchanges occur between the patient,
teacher and trainee. Both the patient-trainee interaction, as well as the teacher-trainee interaction, stimulate the trainee to define problems, synthesize solutions and recognize personal knowledge shortcomings. These cognitive and metacognitive processes result in highly educational experiences.
Overview of MVP:
In order to create a simulation environment that enables these processes, two essential design requirements are needed: the creation of electronic patients; and an electronic forum for communicative exchanges in English to occur between patient and trainee. Both requirements are functional in the MVP.
To create electronic patients, the MVP environment processes encoded deep medical knowledge using artificial intelligence software in a multi-agent network to generate Virtual Patients (VPs) as instances. These VP instances demonstrate realistic physiological and medical findings that evolve over time in an automatic way. The VPs also react appropriately
to trainee chosen medical interventions, whether correct or incorrect. The combination of the depth and complexity of the VP along with the trainee’s actions results in a dynamic, open-ended process where many VP outcomes are possible.
To create an electronic forum for a two-way conversation between VP and trainee, the MVP environment uses natural language processing technology to assign meaning to trainee input. Once the input is interpreted, the VP formulates a response, also in natural language, to return to the trainee. This response is electronically generated through intelligent
reasoning that considers the trainee input request, MVP deep medical knowledge, VP physiological traits, and VP intellectual traits and social values.
A third design component of the MVP is to enable a two-way conversation between a trainee and an electronic teacher. Although not as well developed as the VP discussed above, the intelligent virtual teacher in the MVP is capable of evaluating the trainee’s performance real-time and providing selected feedback and suggestions, also in natural language.
This capability will become more developed as additional knowledge is added to the system.
Technical Accomplishments in the MVP:
MVP technical elements are in various stages of refinement and include:
- A knowledge repository containing medical and pedagogical information which is amenable to machine reasoning
- Software capable of reasoning with the knowledge, interpreting trainee language input to the VP and generating VP language output to the trainee
- A knowledge-based model of the human body in health and disease, and the software to use the model to create a VP and have it function autonomously, over time and in response to trainee interventions
- A similar model of human intellectual traits and social values that are important in health maintenance, disease development and disease management, and that are chosen because they are educationally important for a trainee to learn and manage
- Medical knowledge-based and pedagogically-based models of an intelligent virtual teacher
- Software capable of enabling a conversation-based interaction between the trainee, the virtual patient and the virtual teacher using human language
Currently, we have VPs suffering from nine complex esophageal diseases that progress over time. They behave in a clinically appropriate fashion and possess a complexity that challenges clinical skills at the level of a senior GI fellow. Other possible conditions including TBI and PTSD are being considered for additional development.